tools to measure impacts over households of changes in international prices

Please check the latest version of this presentation on:

http://www.agrodep.org/first-annual-workshop ww

w.a

gro

dep

.org

Tools to Measure Impacts

over Households

of Changes in

International Prices

Presented by:

Miguel RoblesInternational Food Policy Research Institute

AGRODEP Workshop on Analytical Tools for Food Prices

and Price Volatility

June 6-7, 2011 • Dakar, Senegal

Tools to Measure Impacts overHouseholds of Changes In International

P iPrices

AGRODEP MEMBERS’ MEETING AND WORKSHOPJUNE 6 8 2011JUNE 6 ‐8, 2011DAKAR, SENEGAL

Miguel RoblesResearch Fellow, IFPRI

1. Overview1. Overview

1. Welfare impact of changing food prices

a) Analytical framework and methodology

b) Empirical estimation: Bangladesh, Pakistan, Vietnam

2. Online welfare impact simulator

3. Other online tools

1. Overview: Welfare impact of changing food iprices

• We answer the following question: What is the impact on g q phouseholds welfare of changing food prices?

• Microeconomic approach

• Data at the household level• Data at the household level

• We go from household level to higher levels of aggregationWe go from household level to higher levels of aggregation (by region, by expenditure quintile, etc.)

1. Overview: Welfare impact of changing food iprices

• Welfare impact estimates rely on the concept of p y pcompensating variation: amount of extra income required by a household in order to compensate this household for a change in priceschange in prices.

• We take into account the fact that households might be gconsumers and producers of food (key in rural areas)

• Households’ consumption and production decisions respond to price changes (substitution effects)

2. Welfare effects: Methodology2. Welfare effects: Methodology

• We estimate “compensating variation” for allWe estimate compensating variation for all households– A formal representation of compensating variation is derived in Robles

&& Torero 2010

W f h h ld• We compare for every household:– Expenditure “without” price shock (directly estimated from household

surveys)

– Expenditure “with” shock = Expenditure “without” price shock –Compensating variation


• What do we need to estimate compensatingWhat do we need to estimate compensating variation?– Define commodities or group of commodities

– Compute expenditure shares and production shares

– Compensated demand elasticities (own and cross price elasticities)

• We estimate econometrically a system of demand equations, the quadratic almost ideal demand system (QUAIDS):quadratic almost ideal demand system (QUAIDS):


• What do we need to estimate compensatingWhat do we need to estimate compensating variation?

– Define price changes (we simulate full and partial transmission for producers)

– Estimate total household expenditure with and without price shock


• We want to estimate by how much the welfare of household “i” is affected when faced with a change in the price of food.

• Idea is to estimate the change in welfare for any household “i”

Qx

?)(

food

i

dPdU

Ui

Qfood

3. Welfare effects: Methodology3. Welfare effects: Methodology• Graphical representation of concept of compensating variation... It

provides good intuition

Qx

Initial situation at original pricesg pSituation after increase in food

price

Notice that (nominal) budget is the

01

Notice that (nominal) budget is the same but welfare decreases

Qfood

3. Methodology: compensating variation3. Methodology: compensating variation

Qx Budget needed to buy same bundle as in (0) but at new prices

2Budget needed to achieve same welfare as in (0) but at new prices

01

Qfood


Substitution effect

Qx Budget needed to buy same bundle as in (0) but at new prices

2Budget needed to achieve same welfare as in (0) but at new prices

01

Compensating variation

QfoodDirect or “first round” effect


• Graphical representation provides good intuition but it is an incomplete story !!!– We need to deal with several food and non‐food commodities

– We need to incorporate the possibility of food production

• A formal representation of the compensating variation is needed:• A formal representation of the compensating variation is needed:

1, ,

12

1


• What do we need to estimate (1) ?– Define commodities or group of commodities

– Compute expenditure shares and production shares

– Compensated demand elasticities (own and cross price elasticities)

– Price changes (full, partial transmission)

– Total household expenditure

Price changes (%)(vector)

Total expenditure

, ,12

1

Consumption shares(vector)

Demand elasticities(matrix)

Production shares(vector)


• Direct effect and substitution effect

• Elasticities

Direct effect Substitution effect

, ,12

1 2

Elasticities: We estimate econometrically a system of demand equations, the quadratic almost ideal demand system (QUAIDS):

l l l2

21

ln ln1

b ln 2


Elasticities: We estimate econometrically a system of demand equations, the quadratic almost ideal demand system (QUAIDS):

ln ln ln2

21

ln ln1

bln 2

3. Empirical strategy3. Empirical strategy

• Definition of commodity groups

TABLE 1: Commodity groupsBangladesh Pakistan Vietnam

1. Rice 1. Rice 1. Rice

2. Bread and Wheat 2. Bread and Wheat 2. Bread and Wheat

3. Legumes & Pulses 3. Legumes & Pulses 3. Legumes & Pulses

4. Roots & Tubes 4. Roots & Tubes 4. Roots & Tubes

5. Vitamin A‐rich fruits & vegetables 5. Vitamin A‐rich fruits & vegetables 5. Vitamin A‐rich fruits & vegetables

6. Other fruits & vegetables 6. Other fruits & vegetables 6. Other fruits & vegetables

7. Meat, Fish & Dairy 7. Meat, Fish & Dairy 7. Meat, Fish & Dairy

8. Oils & Fats 8. Oils & Fats 8. Oils & Fats

9. Sugars 9. Sugars 9. Sugars

10. Others 10. Others 10. Alcohol

11. Non food 11. Non food 11. Others

12. Non food


• Change in food prices: 3 simulations

TABLE 2: Change in food prices Commodity group dP1 dP2 dP3

Ban Pak Vie All All1 Ri 46 51% 49 75% 38 93% 10 00% 20 00%1. Rice 46.51% 49.75% 38.93% 10.00% 20.00%2. Bread and Wheat 81.48% 32.01% 62.01% 10.00% 20.00%3. Legumes & Pulses 3.00% 7.94% 4.72% 10.00% 20.00%4. Roots & Tubes 3.00% 7.94% 4.72% 10.00% 20.00%5. Vitamin A‐rich fruits & vegetables 3.00% 7.94% 4.72% 10.00% 20.00%6. Other fruits & vegetables 3.00% 7.94% 4.72% 10.00% 20.00%7. Meat, Fish & Dairy 3.00% 7.94% 4.72% 10.00% 20.00%8. Oils & Fats 3.00% 7.94% 4.72% 10.00% 20.00%9. Sugars 3.00% 7.94% 4.72% 10.00% 20.00%10 Alcohol ‐‐ ‐‐ 4 72% 10 00% 20 00%10. Alcohol 4.72% 10.00% 20.00%10/11. Others 0.00% 0.00% 0.00% 0.00% 0.00%11/12. Non food 0.00% 0.00% 0.00% 0.00% 0.00%

dP1: Real change in observed consumer prices between Q1 2006 and Q2 2008

dp2: flat 10% increase in all food groups other that other food

dp3: flat 20% increase in all food groups other that other food


• Estimation of expenditure “without” and “with” food price shock (we subtract the compensating variation)…. Most people would call it “before” and “after” shock

• Analysis of “poverty dynamics”

HH expenditure Poverty pwith price shock

yline

Non‐poor better

Example:

Poverty line

Povertyexit Non‐poor

worse

pHousehold who is poor “without “ shock and due to price shock exits poverty

line

Povertyentry

Povertydeepening

Povertyalleviation

HH expenditure without price shock

45o

deepening

3. Empirical strategy: scenarios3. Empirical strategy: scenarios

• Scenario A:

Ob d h i f d i– Observed change in food prices

– Consumer prices and farm gate prices increase in the same proportion

– Substitution effects

• Scenario 10%:

– 10% change in food prices

– Consumer prices and farm gate prices increase in the same proportion

– Substitution effects

• Other scenarios considered in the study (not in this presentation)

– No change in farm gate prices

– Partial change in farm gate pricesg g p

– No substitution effects

– 20% change in food prices

4. DATA4. DATA

Bangladesh Pakistan Vietnam

SurveyHousehold Income and

E ditPakistan Social and Living St d d M t

Household Living St d d S

yExpenditure Standards Measurement Standards Survey

Year 2005 2005‐06 2006

Sample size (HHs) 10,080 15,4539,189

Rural (%) 63.5 59.6 74.9Urban (%) 36 5 40 4 25 1Urban (%) 36.5 40.4 25.1

5. Results5. Results

• Here we focus on – Scenario A: estimated real food price changes between Q1 2006 and Q1 2008

– Scenario 10%: common 10% price shock to food prices in all three countries

• Analysis on:– Proportion of losers (and winners)

– Size of losses

– Aggregate loss

– Poverty Dynamics

• Basic background information:– Bangladesh: 1,399 US$ GDP pc (PPP) Ranking = 154

– Pakistan: 2,624 US$ GDP pc (PPP) Ranking = 133, p ( ) g

– Vietnam 2,794 US$ GDP pc (PPP) Ranking = 129

5. Results: Scenario A5. Results: Scenario A

• Urban areas are the big losers

• But also more than 70% of rural households are worse off in Bangladesh and Pakistan

• Vietnam is different, 2/3 of rural households are better off and even more in the poorest quintile…

TABLE #: Proportion of losers by expenditure quintile(Scenario A, %)


Quintile Urban Rural National Urban Rural National Urban Rural National

1 93.9 84.4 85.3 96.9 79.3 81.4 70.5 16.0 17.8

2 94 4 76 6 77 5 98 4 74 5 79 3 89 4 25 7 33 72 94.4 76.6 77.5 98.4 74.5 79.3 89.4 25.7 33.73 96.6 72.9 75.9 97.7 71.0 79.1 94.9 36.4 51.64 95.6 68.6 77.9 97.8 68.3 78.9 96.3 45.1 67.85 96.3 72.1 84.1 98.2 68.5 87.1 97.6 56.0 82.1

Total 95.4 74.9 80.1 97.8 72.3 81.1 89.7 35.8 50.6


• Regressive effect !!! When we look at loser households in all 3 countries the poor suffer more (relative to their expenditure level)…

TABLE XX M ti i ti i A (% f h h ld dit )TABLE XX: Mean compensating variation scenario A (% of household expenditure) (Only losers included)



1 14.0 16.0 15.6 6.4 7.4 7.1 6.3 8.2 8.1

2 10.8 13.2 12.7 5.3 6.5 6.2 4.6 7.2 6.8

3 8.7 11.4 10.8 4.7 6.0 5.5 3.9 5.9 5.3

4 7.0 9.4 8.8 3.8 5.4 4.9 3.0 4.7 4.0

5 4.4 7.4 6.2 2.5 4.5 3.3 2.1 2.9 2.6

Total 8 9 11 7 10 8 4 5 6 0 5 4 3 8 5 1 4 5Total 8.9 11.7 10.8 4.5 6.0 5.4 3.8 5.1 4.5


• The aggregate losses are largest in Bangladesh, almost 7% of national consumption expenditure.

I B l d h i ld b i f ll h b 40% 2 54% f i l i• In Bangladesh it would be expensive to fully compensate the bottom 40%: 2.54% of national consumption expenditure is required.

• In Vietnam only 0.31% is required.

TABLE #: Compensating variation (CV) as % of national consumption expenditure(Scenario A, Only losses are included)

Bangladesh Pakistan VietnamBangladesh Pakistan VietnamQuintile Urban Rural National Urban Rural National Urban Rural National

1 0.39 0.93 1.28 0.32 0.41 0.67 0.15 0.05 0.092 0 42 0 91 1 26 0 34 0 42 0 72 0 21 0 11 0 222 0.42 0.91 1.26 0.34 0.42 0.72 0.21 0.11 0.223 0.45 0.87 1.24 0.34 0.42 0.74 0.25 0.17 0.374 0.47 0.84 1.39 0.34 0.40 0.76 0.25 0.23 0.565 0.55 1.05 1.73 0.37 0.42 0.88 0.28 0.32 0.79

Total 2.29 4.61 6.90 1.70 2.07 3.77 1.14 0.89 2.03


• In Bangladesh and Pakistan poverty rates increase, especially in urban areas

A d i b h i d 80% f h l ff• And in both countries around 80% of the rural poor are worse off

• In Vietnam there is an important positive effect on poverty (national poverty rate decreases 7.8% percent points)…

TABLE XX: Poverty dynamics scenario A (% of households)


Urban Rural National Urban Rural National Urban Rural National

Poverty Deepening 29.2% 36.5% 34.7% 13.1% 25.9% 21.5% 1.7% 4.9% 4.0%

Poverty Alleviation 1.3% 5.5% 4.4% 0.4% 4.4% 3.0% 1.0% 11.4% 8.6%

Poverty Exit 0.4% 3.3% 2.6% 0.2% 3.0% 2.0% 0.9% 12.0% 9.0%

Poverty Entry 7 6% 7 9% 7 8% 3 4% 4 3% 4 0% 0 7% 1 3% 1 1%Poverty Entry 7.6% 7.9% 7.8% 3.4% 4.3% 4.0% 0.7% 1.3% 1.1%

Non poor worse 58.5% 30.6% 37.7% 81.3% 42.1% 55.7% 87.3% 29.6% 45.5%

Non poor better 3.0% 16.3% 12.9% 1.7% 20.3% 13.8% 8.3% 40.8% 31.9%

100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

Poverty change 7.2% 4.6% 5.3% 3.3% 1.3% 2.0% ‐0.2% ‐10.7% ‐7.8%

5. Results: Scenario C5. Results: Scenario C

• But the positive effect in Vietnam happens under the assumption of price transmission to producers

If li i i i i d h i Vi i b 1 5 i• If we eliminate price transmission to producers then poverty in Vietnam increases by 1.5 percent points

• Negative effects are magnified in all countries

TABLE XX Poverty dynamics scenario C (% of households)TABLE XX: Poverty dynamics scenario C (% of households)





Poverty Exit 0.0% 0.1% 0.1% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%

Poverty Entry 7.7% 8.1% 8.0% 3.5% 5.2% 4.6% 0.8% 1.8% 1.5%

Non poor worse 60.6% 41.3% 46.2% 82.9% 61.5% 68.9% 95.3% 67.6% 75.2%

Non poor better 0.9% 5.3% 4.2% 0.0% 0.0% 0.0% 0.3% 2.3% 1.8%

100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

Poverty change 7.7% 8.0% 8.0% 3.5% 5.2% 4.6% 0.8% 1.8% 1.5%

5. Results: Cross‐country comparison (10% shock)y p ( )

• In terms of proportion of winner and losers we get similar results:– Great majority if not all urban households are worse off in all 3 countries

– Also true for rural Bangladesh and Pakistan (more than 80%) but different story in Vietnam

– Vietnam: 60 % of rural households become winners– Vietnam: 60 % of rural households become winners

• We confirm regressive effect

O l h h ld ff d i i l di l• On average loser households suffer a reduction in real expenditure equal to:

– Bangladesh = 4.8%

Pakistan = 3 8%– Pakistan = 3.8%

– Vietnam = 3.1%


• We also confirm that the size of aggregate losses are relatively large in Bangladesh (3.95%) and Pakistan (3 02%) and lower in Vietnam (1 69%)(3.02%) and lower in Vietnam (1.69%)

• Vietnam is different because full compensation to the bottom 40% is much lower than in Bangladesh and Pakistan

– Bangladesh = 1.09%, Pakistan = 1%, Vietnam = 0.18%

TABLE #: Compensating variation (CV) as % of national consumption expenditure

(Scenario 10%, Only losses are included)



1 0 17 0 36 0 50 0 22 0 28 0 46 0 10 0 02 0 051 0.17 0.36 0.50 0.22 0.28 0.46 0.10 0.02 0.05

2 0.22 0.42 0.59 0.25 0.31 0.54 0.16 0.07 0.13

3 0.27 0.46 0.68 0.27 0.33 0.59 0.22 0.11 0.27

4 0.32 0.52 0.86 0.28 0.35 0.64 0.24 0.18 0.46

5 0.45 0.77 1.31 0.34 0.39 0.80 0.31 0.28 0.78

Total 1.42 2.53 3.95 1.36 1.66 3.02 1.04 0.66 1.69


• The largest increase in poverty rates takes place in Bangladesh (+3.8%). In Pakistan (+2.2%)

• In Vietnam poverty is reduced (‐1.8%)

• In Vietnam only 20% of poor rural households are worse off (high poverty alleviation effect)

TABLE XX: Poverty dynamics scenario 10% (% of households)





Poverty Exit 0.1% 0.5% 0.4% 0.0% 1.0% 0.7% 0.4% 3.3% 2.5%

Poverty Entry 3.8% 4.3% 4.2% 2.2% 3.2% 2.9% 0.4% 0.8% 0.7%

Non poor worse 64.4% 43.3% 48.7% 83.0% 50.8% 61.9% 88.1% 33.2% 48.3%

Non poor better 0.9% 7.2% 5.6% 1.2% 12.6% 8.7% 7.8% 37.8% 29.6%

100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0%

Poverty change 3.7% 3.8% 3.8% 2.2% 2.2% 2.2% 0.0% ‐2.5% ‐1.8%

6. Final Comments

• Our findings show that the welfare impact of changing food prices is more or less similar in B l d h d P ki h i Vi i ff diff U i li dBangladesh and Pakistan; however, in Vietnam impact effects are different. Using realized price (2006‐2008) shocks we find:

• The fraction of households that are worse off due to increasing food prices is very similar in g p yBangladesh and Pakistan, around 80 percent. In Vietnam, this fraction is much smaller, 50.6 percent

• In urban areas the great majority of households suffer losses due to increasing food prices as• In urban areas the great majority of households suffer losses due to increasing food prices, as they do not engage in food production. Hence, most positively impacted households are located in rural areas. However, even in this region, in Bangladesh and Pakistan three‐fourths of the rural households do not benefit from higher food prices. On the contrary, in Vietnam, 64 2 percent of rural households are better off64.2 percent of rural households are better off.

• Among negatively impacted households there is a clear regressive effect pattern. This regressive pattern is more evident in urban areas

6. Final Comments

• The largest average losses (again as fraction of household expenditure) are observed in B l d h (10 8 ) f ll d b P ki (5 4 ) I Vi h l fBangladesh (10.8 percent), followed by Pakistan (5.4 percent). In, Vietnam the average loss of negatively impacted households is 4.5 percent.

• The size of aggregate losses measured as a fraction of national aggregate consumption gg g gg g pexpenditure is also largest in Bangladesh (6.9 percent), followed by Pakistan (3.8 percent). In Vietnam, this number is smaller (2 percent).

• The total cost of fully compensating the losses of the poorest households (bottom quintile) is• The total cost of fully compensating the losses of the poorest households (bottom quintile) is far from small in Bangladesh (1.3 percent of national aggregate expenditure) and about half in Pakistan (0.67 percent of national aggregate expenditure). In Vietnam, this cost is relatively low (0.31 percent of national aggregate expenditure).

• Poverty rates increase in Bangladesh and Pakistan (5.3 and 2 percent) but decrease in Vietnam (‐7.8 percent). However, if we eliminate the price transmission channel to food producers, the poverty in Vietnam also increases (1.5 percent). This points to the importance of having reliable information at disaggregated levels on how food prices change and whether farmers really can benefit from higher food prices.

6. Final Comments

When we use a common price shock across countries (10 and 20 percent increase in most food items), we observe the following results:observe the following results:

• Again we find similarities between Bangladesh and Pakistan but differences in Vietnam. Regardless of a 10 or 20 percent price shock, the proportion of negatively impacted households in Bangladesh is very close to 9 percent while in Pakistan it is slightly more than 12 percent In Vietnam 46 percent of all households9 percent, while in Pakistan it is slightly more than 12 percent. In Vietnam, 46 percent of all households benefit from higher food prices. In rural areas, this proportion is as high as 60 percent.

• Total aggregate losses (measured as a fraction of total national aggregate consumption expenditure) are not that different in Bangladesh and Pakistan—3.95 percent in the former and 3.02 in the latter. In g pVietnam, given the much larger proportion of positively impacted households, the total aggregate loss only reaches 1.7 of aggregate expenditure.

• While in Bangladesh and Vietnam a 10 percent price shock increases the national poverty rates, in Vietnam the poverty rate decreases. The largest effect on poverty rate happens in Bangladesh; here, poverty rate increases by 3.8 percent. In Pakistan, the increase is equal to 2.2 percent. In Vietnam, the effects on poverty are completely different. A 10 price shock reduces rural poverty by 2.5 percent, while urban poverty remains unchanged.

II. Online welfare impact simulator

Welfare impact of changing food prices: online simulator

Welfare impact of changing food prices: online simulator(developer testing site)

User‐defined

1. Single household Descriptive

parameters

1. Single household simulations statistics

Food consumption patterns

What is the welfare impact of changing

Define Food price

changes2. Country level simulations

patterns

Welfare impact

food prices?changes

Poverty impact

Country representative

household surveys

1. Single household simulations

Descriptive statisticsWhat is the

welfare Define Food

User‐defined parameters

2. Country level simulations


impact of changing

food prices?

Food price

changes

U d fi d tUser defined‐parameters



welfare Define Food




impact of changing

food prices?

Food price

changes

• Option to simulate consumption and productionconsumption and production elasticities•This captures substitutions effects

Simulation results




welfare impact of

Define Food



impact of changing

food prices?

price changes

Welfare impact


household surveys

Poverty impact

Current programming development stage: GuatemalaInitial set of countries for which data is available:

•Latin America: Mexico, Guatemala, Honduras, El Salvador, Nicaragua, Costa Rica Panama Dominican Republic Jamaica Ecuador PeruCosta Rica, Panama, Dominican Republic, Jamaica, Ecuador, Peru•Africa: Kenya, Ghana•Asia: Bangladesh, Pakistan, Vietnam

Additional datasets can enter the system over time

simulations


statistics


Welfare impact

What is the welfare impact of changing

food prices?

Define Food price

changes

p

Poverty impact


household surveys

THANK YOU !

tools to measure impacts over households of changes in international prices

Education